Abstract

Searching of an image is a technique to find the required image from the collection of images. This paper deal with the confront the trouble of automatic generation of caption for news images which are collocated with thematically related documents as well as development of efficient tools that generate description for images automatically which is more advantageous to image search engines that get welfare from image description in supporting more accurate and targeted queries for end users. This paper deals with the generation of captions from the database of images, news articles that captures the image’s contents and consist of two factors that are content selection and surface realization. Content selection shows relationship between appearance of certain features in a documents with the appearance of corresponding features in a given images whereas surface realization arbitrate verbalization of the chosen contents. Annotation process applies over the images and documents collection available in the database. To render the extracted image content in natural language without relying on rich knowledge resources, sentence-templates or grammars we need to be considering both extractive and abstractive caption generation models. In the paper we will scrutinize phrase based model for caption generation as a consequence with phrase tree construction. After making comparisons between abstractive and extractive method it is examine that output of abstractive model is better than extractive method.

Paper Details